[BUG]: Perplexity model name does not match models available via the API

[rahul-tar]

How are you running VectorAdmin?

Docker (local)

What happened?

The model name under perplexity is wrong and does not match any model on perplexity model card.

Are there known steps to reproduce?

Follow these steps:
Chat Settings > Workspace LLM Provider > Perplexity AI > Workspace Chat Model > llama-3.1-sonar-huge-128k-chat

Now try chatting and it will say invalid model. That's because llama-3.1-sonar-huge-128k-chat is not the right name for the 405B model. It is llama-3.1-sonar-huge-128k-online as seen in docs here - https://docs.perplexity.ai/guides/model-cards

[timothycarambat]

This chore is so annoying. I really wish they would just host a /models endpoint so we dont have to keep doing this...

[rahul-tar]

This chore is so annoying. I really wish they would just host a /models endpoint so we dont have to keep doing this...

Totally get that!
Maybe scrape their model cards URL periodically to get model names?

[timothycarambat]

Would be way more effort than its worth because they also keep changing the docs page too! 😆

~$3B company and can't make a /models endpoint like everyone else, crazy!

[rahul-tar]

Would be way more effort than its worth because they also keep changing the docs page too! 😆

~$3B company and can't make a /models endpoint like everyone else, crazy!

brainstorming here: Maybe a feature where user can type in model name in addition to the preset drop down?

[timothycarambat]

Ideally, yeah - basically a combobox for those who dont allow us to pull models dynamically!